Gst_nvstreammux plugins segmentation fault

Hi everyone:
I am working with Jetson NX and Jetpack version 4.3,deepstream 5.0, tensor 7.1.3-1+cuda10.2. The app modified based on the deepstream-test3. when i run the sample “./deepstream-test3-app file:///home/ubuntu/video1.mp4” , I get the segmentation fault:

e-engine> [UID 1]: Load new model:dstest3_pgie_config.txt sucessfully
Decodebin child added: source
Decodebin child added: decodebin0
[New Thread 0x7f6d8551f0 (LWP 19206)]
Decodebin child added: qtdemux0
[New Thread 0x7f6d0541f0 (LWP 19207)]
Decodebin child added: multiqueue0
[New Thread 0x7f6c8531f0 (LWP 19208)]
Decodebin child added: h264parse0
Decodebin child added: capsfilter0
Decodebin child added: nvv4l2decoder0
NvMMLiteOpen : Block : BlockType = 261
NVMEDIA: Reading vendor.tegra.display-size : status: 6
[New Thread 0x7f677fe1f0 (LWP 19209)]
[New Thread 0x7f66ffd1f0 (LWP 19210)]
[New Thread 0x7f667fc1f0 (LWP 19211)]
[New Thread 0x7f65ffb1f0 (LWP 19212)]
NvMMLiteBlockCreate : Block : BlockType = 261
[New Thread 0x7f657fa1f0 (LWP 19213)]
In cb_newpad
[New Thread 0x7f64ff91f0 (LWP 19214)]
[New Thread 0x7f4ffff1f0 (LWP 19216)]
[New Thread 0x7f4f7fe1f0 (LWP 19217)]
[New Thread 0x7f4effd1f0 (LWP 19218)]
[New Thread 0x7f4e7fc1f0 (LWP 19219)]
[New Thread 0x7f3b7ff1f0 (LWP 19220)]

Thread 18 “stream-muxer-1:” received signal SIGSEGV, Segmentation fault.
[Switching to Thread 0x7f4ffff1f0 (LWP 19216)]
0x0000000200000000 in ?? ()
(gdb) bt
#0 0x0000000200000000 in ()
#1 0x0000005555b444b0 in ()

how can i debug it.

The attachment i s my code.deepstream_test3_app.c (32.9 KB)

Please do not use videorate after decode bin, videorate can not work with deepstream plugins.

The videerate is removed,but the problem persists. I have update the lib as the follow link , Nvmultistreamtiler - Segmentation fault, but this bug is not fixed.

It is not reasonable to ask us to debug your code unless you can prove that it is a deepstream problem(either you find the problem is caused by some deepstream plugin or you find the problem is caused by some deepstream configuration). Can you write the “gst-launch-1.0” command to prove your pipeline in the attached cpp file is correct? Or you can dump the gstreamer graph file for analysis, the graph dump method is in DeepStream SDK FAQ - Intelligent Video Analytics / DeepStream SDK - NVIDIA Developer Forums.

After a glimpse to your code, the nvmultistreamtiler should not be put after nvdsosd, please refer to the original deepstream-test3 sample code.

And we have a known bug for nvstreamdemux with RGBA format. DeepStream 5.0.1 nvstreamdemux core dump after RGBA conversion - Intelligent Video Analytics / DeepStream SDK - NVIDIA Developer Forums, please avoid this in your pipeline.

1 Like

OK , thank you for you reply, I will try the way as you suggest.